Check working directory

getwd()
## [1] "/Users/alexg/R files/hair_cortisol/skew-normal FINAL"

Load packages

library(readxl)
library(psych)
library(dlookr)
## Registered S3 methods overwritten by 'dlookr':
##   method          from  
##   plot.transform  scales
##   print.transform scales
## 
## Attaching package: 'dlookr'
## The following object is masked from 'package:psych':
## 
##     describe
## The following object is masked from 'package:base':
## 
##     transform
library(vtable)
## Loading required package: kableExtra
library(dplyr)
## 
## Attaching package: 'dplyr'
## The following object is masked from 'package:kableExtra':
## 
##     group_rows
## The following objects are masked from 'package:stats':
## 
##     filter, lag
## The following objects are masked from 'package:base':
## 
##     intersect, setdiff, setequal, union
library(reshape)
## 
## Attaching package: 'reshape'
## The following object is masked from 'package:dplyr':
## 
##     rename
library(ggplot2)
## 
## Attaching package: 'ggplot2'
## The following objects are masked from 'package:psych':
## 
##     %+%, alpha
library(brms)
## Loading required package: Rcpp
## Loading 'brms' package (version 2.22.0). Useful instructions
## can be found by typing help('brms'). A more detailed introduction
## to the package is available through vignette('brms_overview').
## 
## Attaching package: 'brms'
## The following object is masked from 'package:psych':
## 
##     cs
## The following object is masked from 'package:stats':
## 
##     ar
library(rethinking)
## Loading required package: cmdstanr
## This is cmdstanr version 0.8.0
## - CmdStanR documentation and vignettes: mc-stan.org/cmdstanr
## - CmdStan path: /Users/alexg/.cmdstan/cmdstan-2.36.0
## - CmdStan version: 2.36.0
## Loading required package: posterior
## This is posterior version 1.6.1
## 
## Attaching package: 'posterior'
## The following object is masked from 'package:dlookr':
## 
##     entropy
## The following objects are masked from 'package:stats':
## 
##     mad, sd, var
## The following objects are masked from 'package:base':
## 
##     %in%, match
## Loading required package: parallel
## rethinking (Version 2.42)
## 
## Attaching package: 'rethinking'
## The following objects are masked from 'package:brms':
## 
##     LOO, stancode, WAIC
## The following objects are masked from 'package:psych':
## 
##     logistic, logit, sim
## The following object is masked from 'package:stats':
## 
##     rstudent
library(loo)
## This is loo version 2.8.0
## - Online documentation and vignettes at mc-stan.org/loo
## - As of v2.0.0 loo defaults to 1 core but we recommend using as many as possible. Use the 'cores' argument or set options(mc.cores = NUM_CORES) for an entire session.
## 
## Attaching package: 'loo'
## The following object is masked from 'package:rethinking':
## 
##     compare
library(priorsense)
library(tidyverse)
## ── Attaching core tidyverse packages ──────────────────────── tidyverse 2.0.0 ──
## ✔ forcats   1.0.0     ✔ stringr   1.5.1
## ✔ lubridate 1.9.4     ✔ tibble    3.2.1
## ✔ purrr     1.0.4     ✔ tidyr     1.3.1
## ✔ readr     2.1.5
## ── Conflicts ────────────────────────────────────────── tidyverse_conflicts() ──
## ✖ ggplot2::%+%()      masks psych::%+%()
## ✖ ggplot2::alpha()    masks psych::alpha()
## ✖ tidyr::expand()     masks reshape::expand()
## ✖ tidyr::extract()    masks dlookr::extract()
## ✖ dplyr::filter()     masks stats::filter()
## ✖ dplyr::group_rows() masks kableExtra::group_rows()
## ✖ dplyr::lag()        masks stats::lag()
## ✖ purrr::map()        masks rethinking::map()
## ✖ reshape::rename()   masks dplyr::rename()
## ✖ lubridate::stamp()  masks reshape::stamp()
## ℹ Use the conflicted package (<http://conflicted.r-lib.org/>) to force all conflicts to become errors
library(vioplot)
## Loading required package: sm
## Package 'sm', version 2.2-6.0: type help(sm) for summary information
## 
## Attaching package: 'sm'
## 
## The following object is masked from 'package:dlookr':
## 
##     binning
## 
## Loading required package: zoo
## 
## Attaching package: 'zoo'
## 
## The following objects are masked from 'package:base':
## 
##     as.Date, as.Date.numeric
library(bayesplot)
## This is bayesplot version 1.12.0
## - Online documentation and vignettes at mc-stan.org/bayesplot
## - bayesplot theme set to bayesplot::theme_default()
##    * Does _not_ affect other ggplot2 plots
##    * See ?bayesplot_theme_set for details on theme setting
## 
## Attaching package: 'bayesplot'
## 
## The following object is masked from 'package:posterior':
## 
##     rhat
## 
## The following object is masked from 'package:brms':
## 
##     rhat
library(bayestestR)

Load data

df <- read_xlsx("hair_cort_dog_all.xlsx", col_types = c("text", "text",  
                               "text", "text", "text", "text",
                               "text", "numeric","text", "skip",
                               "numeric", "skip", "skip", 
                               "numeric", "skip"))
df <- as.data.frame(df)

INITIAL DATA PLOTTING AND EXPLORATION

Check characteristics of df

dim(df) # will tell you how many rows and columns the dataset has
## [1] 73 11
class(df) # will tell you what data structure has the dataset been assigned
## [1] "data.frame"

Explore the dataset to understand its structure.

head(df)
##   number   group visit season breed_group coat_colour    sex age comorbidity
## 1     c1 stopped    v0 winter         ret        dark   Male  43         yes
## 2     c2 stopped    v0 autumn         mix        dark   Male 105         yes
## 3     c3 stopped    v0 spring        ckcs         mix Female 117         yes
## 4     c4 stopped    v0 summer         ret        dark Female 108         yes
## 5     c5 stopped    v0 summer         ret        dark Female 110         yes
## 6     c6 stopped    v0 winter         mix       light Female 120         yes
##   fat_percent cortisol
## 1    52.21393 4.924220
## 2    38.52059 7.304202
## 3    46.94916 1.590000
## 4    44.46813 0.861570
## 5    39.59363 6.217317
## 6          NA 4.426785

1. Get summary stats for numeric data

numeric_df <- Filter(is.numeric, df)
describe(numeric_df) # the describe function in psych provides summary stats
## # A tibble: 3 × 26
##   described_variables     n    na  mean    sd se_mean   IQR skewness kurtosis
##   <chr>               <int> <int> <dbl> <dbl>   <dbl> <dbl>    <dbl>    <dbl>
## 1 age                    73     0 95.8  35.6     4.16 44      -0.104 -0.00589
## 2 fat_percent            55    18 40.5   7.82    1.05  7.82   -0.294  1.12   
## 3 cortisol               73     0  8.11 16.5     1.93  5.43    4.05  18.7    
## # ℹ 17 more variables: p00 <dbl>, p01 <dbl>, p05 <dbl>, p10 <dbl>, p20 <dbl>,
## #   p25 <dbl>, p30 <dbl>, p40 <dbl>, p50 <dbl>, p60 <dbl>, p70 <dbl>,
## #   p75 <dbl>, p80 <dbl>, p90 <dbl>, p95 <dbl>, p99 <dbl>, p100 <dbl>

2. Check normality of all numeric variables

a. graphical assessment

plot_normality(numeric_df)

b. shapiro-wilk test

apply(numeric_df, 2, shapiro.test)
## $age
## 
##  Shapiro-Wilk normality test
## 
## data:  newX[, i]
## W = 0.97361, p-value = 0.1288
## 
## 
## $fat_percent
## 
##  Shapiro-Wilk normality test
## 
## data:  newX[, i]
## W = 0.97956, p-value = 0.4692
## 
## 
## $cortisol
## 
##  Shapiro-Wilk normality test
## 
## data:  newX[, i]
## W = 0.46269, p-value = 6.756e-15

c. repeat Q-Q plots with transformed data

i. log(cortisol)

qqnorm(df$cortisol)
qqline(df$cortisol, col = "red")

qqnorm(log(df$cortisol))
qqline(log(df$cortisol), col = "red")

ii Shapiro test for log cortisol

shapiro.test(log(df$cortisol))
## 
##  Shapiro-Wilk normality test
## 
## data:  log(df$cortisol)
## W = 0.94725, p-value = 0.004126

3. Check data numerically

summary(df$cortisol)
##     Min.  1st Qu.   Median     Mean  3rd Qu.     Max. 
##   0.4141   1.4119   2.3331   8.1089   6.8455 104.6172
summary(log(df$cortisol))
##    Min. 1st Qu.  Median    Mean 3rd Qu.    Max. 
## -0.8817  0.3449  0.8472  1.1816  1.9236  4.6503

a. log-transform cortisol

df$lgCort <- log(df$cortisol)
summary(df$lgCort)
##    Min. 1st Qu.  Median    Mean 3rd Qu.    Max. 
## -0.8817  0.3449  0.8472  1.1816  1.9236  4.6503

i. visualise

hist(df$lgCort)

b. Create simple category name for breed and convert to factor

df$breed <- df$breed_group
df$breed <- factor(df$breed, levels = c("mix", "ckcs", "pug", "ret", "other"))
head(df$breed)
## [1] ret  mix  ckcs ret  ret  mix 
## Levels: mix ckcs pug ret other

4. Generate data summary

sumtable(df)
Summary Statistics
Variable N Mean Std. Dev. Min Pctl. 25 Pctl. 75 Max
group 73
… completed 42 58%
… stopped 31 42%
visit 73
… v0 52 71%
… v1 21 29%
season 73
… autumn 21 29%
… spring 14 19%
… summer 22 30%
… winter 16 22%
breed_group 73
… ckcs 7 10%
… mix 16 22%
… other 26 36%
… pug 7 10%
… ret 17 23%
coat_colour 73
… dark 30 41%
… light 28 38%
… mix 15 21%
sex 73
… Female 43 59%
… Male 30 41%
age 73 96 36 16 73 117 182
comorbidity 73
… no 15 21%
… yes 58 79%
fat_percent 55 40 7.8 18 37 45 61
cortisol 73 8.1 16 0.41 1.4 6.8 105
lgCort 73 1.2 1.2 -0.88 0.34 1.9 4.7
breed 73
… mix 16 22%
… ckcs 7 10%
… pug 7 10%
… ret 17 23%
… other 26 36%

5. Visualise associations

a. between lgCortisol and breed with violin plot

par(mfrow = c(1,1))
vioplot(lgCort ~ breed, col = "firebrick",
        data = df)

b. between lgCortisol and breed with stripchart

stripchart(lgCort ~ breed, vertical = TRUE, method = "jitter",
           col = "steelblue3", data = df, pch = 20)

STANDARDISE DATA FOR MODELLING

1. Standardise cortisol

df$slgCort <- standardize(df$lgC)
summary(df$slgCort)
##    Min. 1st Qu.  Median    Mean 3rd Qu.    Max. 
## -1.7079 -0.6925 -0.2768  0.0000  0.6142  2.8713

a. visualise standardised lgCort

hist(df$slgCort)

2. create dataset only containing complete data

df2 <- na.omit(df)

MODEL FOR THE EFFECT OF BREED ON HAIR CORTISOL

1. Model code

model <- brm(slgCort ~ breed + (1 | visit), family = skew_normal(), data = df2)

2. Check what priors need to be set

default_prior(slgCort ~ breed + (1 | visit),
                   family = skew_normal(),
                   data = df2)
##                    prior     class       coef group resp dpar nlpar lb ub
##             normal(0, 4)     alpha                                       
##                   (flat)         b                                       
##                   (flat)         b  breedckcs                            
##                   (flat)         b breedother                            
##                   (flat)         b   breedpug                            
##                   (flat)         b   breedret                            
##  student_t(3, -0.1, 2.5) Intercept                                       
##     student_t(3, 0, 2.5)        sd                                   0   
##     student_t(3, 0, 2.5)        sd            visit                  0   
##     student_t(3, 0, 2.5)        sd  Intercept visit                  0   
##     student_t(3, 0, 2.5)     sigma                                   0   
##        source
##       default
##       default
##  (vectorized)
##  (vectorized)
##  (vectorized)
##  (vectorized)
##       default
##       default
##  (vectorized)
##  (vectorized)
##       default

Published information about associations with hair cortisol

No published data about effects on breed, but this is plausible However, unclear as to which breeds will differ and which way. Therefore, use a regularising prior but keep it neutral and broad.

3. Set priors

# Set individual priors
prior_int <- set_prior("normal(0, 0.5)", class = "Intercept")
prior_sig <- set_prior("exponential(1)", class = "sigma")
prior_b <- set_prior("normal(0, 1)", class = "b")
prior_sd <- set_prior("normal(0, 1)", class = "sd")
prior_alpha <- set_prior("normal(4, 8)", class = "alpha")

# Combine priors into list
priors <- c(prior_int, prior_sig, prior_b, prior_sd, prior_alpha)

4. Plot priors

a. Prior for intercept

x <- seq(-3, 3, length.out = 100)
y <- dnorm(x, mean = 0, sd = 1)
plot(y ~ x, type = "l")

b. Prior for sigma

x <- seq(0, 3, length.out = 100)
y <- dexp(x, rate = 0.5)
plot(y ~ x, type = "l")

b.ii. deciding on alpha for skew normal distribution

Based on distribution of log normal hair cortisol, expect things to be skewed to the right. Try different levels of alpha for skew normal.

x <- seq(-3, 5, length.out = 100)
y <- dskew_normal(x, mu = 0, sigma = 1, alpha = 4)
plot(y ~ x, type = "l")

So, expect alpha to be positive… perhaps 4, but keep sd broad to allow some flexibility

biii. Prior for visit sd

x <- seq(0, 5, length.out = 100)
y <- dnorm(x, mean = 0, sd = 2)
plot(y ~ x, type = "l")

c. Prior for breed

x <- seq(-3, 3, length.out = 100)
y <- dnorm(x, mean = 0, sd = 1.0)
plot(y ~ x, type = "l")

5. Run model

Increased adapt_delta >0.8 (0.9 here), as had divergent transitions

set.seed(666)
model <- brm(slgCort ~ breed + (1 | visit),
                   family = skew_normal(),
                   data = df,
                   prior = priors,
                   control=list(adapt_delta=0.99999999, stepsize = 0.001, max_treedepth = 15),
                   iter = 8000, warmup = 2000,
                   cores = 4,
                   save_pars = save_pars(all =TRUE),
                   sample_prior = TRUE)
## Compiling Stan program...
## Trying to compile a simple C file
## Running /Library/Frameworks/R.framework/Resources/bin/R CMD SHLIB foo.c
## using C compiler: ‘Apple clang version 17.0.0 (clang-1700.0.13.5)’
## using SDK: ‘MacOSX15.5.sdk’
## clang -arch arm64 -I"/Library/Frameworks/R.framework/Resources/include" -DNDEBUG   -I"/Library/Frameworks/R.framework/Versions/4.4-arm64/Resources/library/Rcpp/include/"  -I"/Library/Frameworks/R.framework/Versions/4.4-arm64/Resources/library/RcppEigen/include/"  -I"/Library/Frameworks/R.framework/Versions/4.4-arm64/Resources/library/RcppEigen/include/unsupported"  -I"/Library/Frameworks/R.framework/Versions/4.4-arm64/Resources/library/BH/include" -I"/Library/Frameworks/R.framework/Versions/4.4-arm64/Resources/library/StanHeaders/include/src/"  -I"/Library/Frameworks/R.framework/Versions/4.4-arm64/Resources/library/StanHeaders/include/"  -I"/Library/Frameworks/R.framework/Versions/4.4-arm64/Resources/library/RcppParallel/include/"  -I"/Library/Frameworks/R.framework/Versions/4.4-arm64/Resources/library/rstan/include" -DEIGEN_NO_DEBUG  -DBOOST_DISABLE_ASSERTS  -DBOOST_PENDING_INTEGER_LOG2_HPP  -DSTAN_THREADS  -DUSE_STANC3 -DSTRICT_R_HEADERS  -DBOOST_PHOENIX_NO_VARIADIC_EXPRESSION  -D_HAS_AUTO_PTR_ETC=0  -include '/Library/Frameworks/R.framework/Versions/4.4-arm64/Resources/library/StanHeaders/include/stan/math/prim/fun/Eigen.hpp'  -D_REENTRANT -DRCPP_PARALLEL_USE_TBB=1   -I/opt/R/arm64/include    -fPIC  -falign-functions=64 -Wall -g -O2  -c foo.c -o foo.o
## In file included from <built-in>:1:
## In file included from /Library/Frameworks/R.framework/Versions/4.4-arm64/Resources/library/StanHeaders/include/stan/math/prim/fun/Eigen.hpp:22:
## In file included from /Library/Frameworks/R.framework/Versions/4.4-arm64/Resources/library/RcppEigen/include/Eigen/Dense:1:
## In file included from /Library/Frameworks/R.framework/Versions/4.4-arm64/Resources/library/RcppEigen/include/Eigen/Core:19:
## /Library/Frameworks/R.framework/Versions/4.4-arm64/Resources/library/RcppEigen/include/Eigen/src/Core/util/Macros.h:679:10: fatal error: 'cmath' file not found
##   679 | #include <cmath>
##       |          ^~~~~~~
## 1 error generated.
## make: *** [foo.o] Error 1
## Start sampling
## Found more than one class "stanfit" in cache; using the first, from namespace 'rethinking'
## Also defined by 'rstan'
## Found more than one class "stanfit" in cache; using the first, from namespace 'rethinking'
## Also defined by 'rstan'
## Found more than one class "stanfit" in cache; using the first, from namespace 'rethinking'
## Also defined by 'rstan'
## Found more than one class "stanfit" in cache; using the first, from namespace 'rethinking'
## Also defined by 'rstan'

6. Get summary of model

summary(model)
##  Family: skew_normal 
##   Links: mu = identity; sigma = identity; alpha = identity 
## Formula: slgCort ~ breed + (1 | visit) 
##    Data: df (Number of observations: 73) 
##   Draws: 4 chains, each with iter = 8000; warmup = 2000; thin = 1;
##          total post-warmup draws = 24000
## 
## Multilevel Hyperparameters:
## ~visit (Number of levels: 2) 
##               Estimate Est.Error l-95% CI u-95% CI Rhat Bulk_ESS Tail_ESS
## sd(Intercept)     0.37      0.36     0.01     1.36 1.00     7556     9552
## 
## Regression Coefficients:
##            Estimate Est.Error l-95% CI u-95% CI Rhat Bulk_ESS Tail_ESS
## Intercept     -0.10      0.30    -0.71     0.51 1.00    10616    11125
## breedckcs      0.14      0.38    -0.65     0.84 1.00    13139    14313
## breedpug       0.19      0.36    -0.55     0.85 1.00    14815    14562
## breedret       0.09      0.28    -0.49     0.62 1.00    12104    15143
## breedother     0.10      0.26    -0.39     0.62 1.00    14704    14362
## 
## Further Distributional Parameters:
##       Estimate Est.Error l-95% CI u-95% CI Rhat Bulk_ESS Tail_ESS
## sigma     1.04      0.10     0.87     1.27 1.00    14360    14487
## alpha     5.35      3.34     1.57    14.44 1.00    10117     6858
## 
## Draws were sampled using sampling(NUTS). For each parameter, Bulk_ESS
## and Tail_ESS are effective sample size measures, and Rhat is the potential
## scale reduction factor on split chains (at convergence, Rhat = 1).

7. MCMC diagnostics

plot(model)

Looking for hairy caterpillars

b. try a trank plot as well

mcmc_plot(model, type = 'rank_overlay')

8. Calculate 95% HPDI for breed

Usually better than the compatability intervals given in the summary ### a. ckcs

draws <- as.matrix(model)
HPDI(draws[,2], 0.97) # 2nd column is draws for ckcs
##      |0.97      0.97| 
## -0.7183886  0.9231527

b. pug

draws <- as.matrix(model)
HPDI(draws[,3], 0.97) # 3rd column is draws for pug
##      |0.97      0.97| 
## -0.5865707  0.9615309

b. ret

draws <- as.matrix(model)
HPDI(draws[,4], 0.97) # 4th column is draws for ret
##      |0.97      0.97| 
## -0.5195171  0.7023931

b. other

draws <- as.matrix(model)
HPDI(draws[,5], 0.97) # 5th column is draws for other
##      |0.97      0.97| 
## -0.4649280  0.6533254

9. Calculate R2 for model

bayes_R2(model, probs = c(0.015, 0.5, 0.985)) # 0.015, 0.5, 0.985 are the quantiles
##      Estimate  Est.Error        Q1.5        Q50     Q98.5
## R2 0.04642802 0.02818366 0.006334857 0.04108256 0.1281878
loo_R2(model, probs = c(0.015, 0.5, 0.985)) # 0.015, 0.5, 0.985 are the quantiles
## Warning: Some Pareto k diagnostic values are too high. See help('pareto-k-diagnostic') for details.
##       Estimate  Est.Error      Q1.5         Q50       Q98.5
## R2 -0.08681619 0.03212178 -0.179703 -0.08211637 -0.03317327

CHECKS ON MODEL

1. Basic check of simulations based on posterior distribution, versus the real data distribution

checks whether actual data is similar to simulated data.

pp_check(model, ndraws = 100) 

2 Check some individual draws versus observed using pp_check

par(mfrow = c(1,1))
pp_check(model, type = "hist", ndraws = 11, binwidth = 0.25) # separate histograms of 11 MCMC draws vs actual data

3. Other pp_check graphs

pp_check(model, type = "error_hist", ndraws = 11) # separate histograms of errors for 11 draws
## `stat_bin()` using `bins = 30`. Pick better value with `binwidth`.

pp_check(model, type = "scatter_avg", ndraws = 100) # scatter plot

pp_check(model, type = "stat_2d") #  scatterplot of joint posteriors
## Using all posterior draws for ppc type 'stat_2d' by default.
## Note: in most cases the default test statistic 'mean' is too weak to detect anything of interest.

# PPC functions for predictive checks based on (approximate) leave-one-out (LOO) cross-validation
pp_check(model, type = "loo_pit_overlay", ndraws = 1000) 
## Warning: Found 2 observations with a pareto_k > 0.7 in model '.x1'. We
## recommend to set 'moment_match = TRUE' in order to perform moment matching for
## problematic observations.
## NOTE: The kernel density estimate assumes continuous observations and is not optimal for discrete observations.

5. Pairs plot

pairs(model)

PSIS LOO-CV to check model performance

loo_model <- loo(model, moment_match = TRUE)
loo_model
## 
## Computed from 24000 by 73 log-likelihood matrix.
## 
##          Estimate   SE
## elpd_loo   -105.1  6.1
## p_loo         6.8  1.6
## looic       210.2 12.2
## ------
## MCSE of elpd_loo is 0.1.
## MCSE and ESS estimates assume MCMC draws (r_eff in [0.4, 1.1]).
## 
## All Pareto k estimates are good (k < 0.7).
## See help('pareto-k-diagnostic') for details.

AUTOMATED PRIOR SENSITIVITY USING THE PRIOR SENSE PACKAGE

1. Sensitivity check

First, check the sensitivity of the prior and likelihood to power-scaling. Posterior and posteriors resulting from power-scaling.

powerscale_sensitivity(model, variable = c("b_Intercept", "b_breedckcs", "b_breedother", "b_breedpug", "b_breedret"), , facet_rows = "variable")
## Sensitivity based on cjs_dist
## Prior selection: all priors
## Likelihood selection: all data
## 
##      variable prior likelihood diagnosis
##   b_Intercept 0.041      0.037         -
##   b_breedckcs 0.025      0.088         -
##  b_breedother 0.019      0.078         -
##    b_breedpug 0.021      0.092         -
##    b_breedret 0.028      0.068         -

2. Kernel density

powerscale_plot_dens(model, variable = c("b_Intercept", "b_breedckcs", "b_breedother", "b_breedpug", "b_breedret"), facet_rows = "variable")

3. Now use bayestestR package to check priors are informative

check_prior(model, effects = "all")
##             Parameter Prior_Quality
## 1         b_Intercept   informative
## 2         b_breedckcs   informative
## 3          b_breedpug   informative
## 4          b_breedret   informative
## 5        b_breedother   informative
## 6 sd_visit__Intercept   informative

CHECK PRIOR PREDICTIVE DISTRIBUTION

1. Prior Predictive Distribution

Can simulate data just on the priors. Fit model but only consider prior when fitting model. If this looks reasonable, it helps to confirm that your priors were reasonable

set.seed(666)
model_priors_only <- brm(slgCort ~ breed +  (1 | visit),
                   family = skew_normal(),
                   prior = priors,
                   data = df,
                   sample_prior = "only")
## Compiling Stan program...
## Trying to compile a simple C file
## Running /Library/Frameworks/R.framework/Resources/bin/R CMD SHLIB foo.c
## using C compiler: ‘Apple clang version 17.0.0 (clang-1700.0.13.5)’
## using SDK: ‘MacOSX15.5.sdk’
## clang -arch arm64 -I"/Library/Frameworks/R.framework/Resources/include" -DNDEBUG   -I"/Library/Frameworks/R.framework/Versions/4.4-arm64/Resources/library/Rcpp/include/"  -I"/Library/Frameworks/R.framework/Versions/4.4-arm64/Resources/library/RcppEigen/include/"  -I"/Library/Frameworks/R.framework/Versions/4.4-arm64/Resources/library/RcppEigen/include/unsupported"  -I"/Library/Frameworks/R.framework/Versions/4.4-arm64/Resources/library/BH/include" -I"/Library/Frameworks/R.framework/Versions/4.4-arm64/Resources/library/StanHeaders/include/src/"  -I"/Library/Frameworks/R.framework/Versions/4.4-arm64/Resources/library/StanHeaders/include/"  -I"/Library/Frameworks/R.framework/Versions/4.4-arm64/Resources/library/RcppParallel/include/"  -I"/Library/Frameworks/R.framework/Versions/4.4-arm64/Resources/library/rstan/include" -DEIGEN_NO_DEBUG  -DBOOST_DISABLE_ASSERTS  -DBOOST_PENDING_INTEGER_LOG2_HPP  -DSTAN_THREADS  -DUSE_STANC3 -DSTRICT_R_HEADERS  -DBOOST_PHOENIX_NO_VARIADIC_EXPRESSION  -D_HAS_AUTO_PTR_ETC=0  -include '/Library/Frameworks/R.framework/Versions/4.4-arm64/Resources/library/StanHeaders/include/stan/math/prim/fun/Eigen.hpp'  -D_REENTRANT -DRCPP_PARALLEL_USE_TBB=1   -I/opt/R/arm64/include    -fPIC  -falign-functions=64 -Wall -g -O2  -c foo.c -o foo.o
## In file included from <built-in>:1:
## In file included from /Library/Frameworks/R.framework/Versions/4.4-arm64/Resources/library/StanHeaders/include/stan/math/prim/fun/Eigen.hpp:22:
## In file included from /Library/Frameworks/R.framework/Versions/4.4-arm64/Resources/library/RcppEigen/include/Eigen/Dense:1:
## In file included from /Library/Frameworks/R.framework/Versions/4.4-arm64/Resources/library/RcppEigen/include/Eigen/Core:19:
## /Library/Frameworks/R.framework/Versions/4.4-arm64/Resources/library/RcppEigen/include/Eigen/src/Core/util/Macros.h:679:10: fatal error: 'cmath' file not found
##   679 | #include <cmath>
##       |          ^~~~~~~
## 1 error generated.
## make: *** [foo.o] Error 1
## Start sampling
## 
## SAMPLING FOR MODEL 'anon_model' NOW (CHAIN 1).
## Chain 1: 
## Chain 1: Gradient evaluation took 6.1e-05 seconds
## Chain 1: 1000 transitions using 10 leapfrog steps per transition would take 0.61 seconds.
## Chain 1: Adjust your expectations accordingly!
## Chain 1: 
## Chain 1: 
## Chain 1: Iteration:    1 / 2000 [  0%]  (Warmup)
## Chain 1: Iteration:  200 / 2000 [ 10%]  (Warmup)
## Chain 1: Iteration:  400 / 2000 [ 20%]  (Warmup)
## Chain 1: Iteration:  600 / 2000 [ 30%]  (Warmup)
## Chain 1: Iteration:  800 / 2000 [ 40%]  (Warmup)
## Chain 1: Iteration: 1000 / 2000 [ 50%]  (Warmup)
## Chain 1: Iteration: 1001 / 2000 [ 50%]  (Sampling)
## Chain 1: Iteration: 1200 / 2000 [ 60%]  (Sampling)
## Chain 1: Iteration: 1400 / 2000 [ 70%]  (Sampling)
## Chain 1: Iteration: 1600 / 2000 [ 80%]  (Sampling)
## Chain 1: Iteration: 1800 / 2000 [ 90%]  (Sampling)
## Chain 1: Iteration: 2000 / 2000 [100%]  (Sampling)
## Chain 1: 
## Chain 1:  Elapsed Time: 0.02 seconds (Warm-up)
## Chain 1:                0.015 seconds (Sampling)
## Chain 1:                0.035 seconds (Total)
## Chain 1: 
## 
## SAMPLING FOR MODEL 'anon_model' NOW (CHAIN 2).
## Chain 2: 
## Chain 2: Gradient evaluation took 2e-06 seconds
## Chain 2: 1000 transitions using 10 leapfrog steps per transition would take 0.02 seconds.
## Chain 2: Adjust your expectations accordingly!
## Chain 2: 
## Chain 2: 
## Chain 2: Iteration:    1 / 2000 [  0%]  (Warmup)
## Chain 2: Iteration:  200 / 2000 [ 10%]  (Warmup)
## Chain 2: Iteration:  400 / 2000 [ 20%]  (Warmup)
## Chain 2: Iteration:  600 / 2000 [ 30%]  (Warmup)
## Chain 2: Iteration:  800 / 2000 [ 40%]  (Warmup)
## Chain 2: Iteration: 1000 / 2000 [ 50%]  (Warmup)
## Chain 2: Iteration: 1001 / 2000 [ 50%]  (Sampling)
## Chain 2: Iteration: 1200 / 2000 [ 60%]  (Sampling)
## Chain 2: Iteration: 1400 / 2000 [ 70%]  (Sampling)
## Chain 2: Iteration: 1600 / 2000 [ 80%]  (Sampling)
## Chain 2: Iteration: 1800 / 2000 [ 90%]  (Sampling)
## Chain 2: Iteration: 2000 / 2000 [100%]  (Sampling)
## Chain 2: 
## Chain 2:  Elapsed Time: 0.02 seconds (Warm-up)
## Chain 2:                0.017 seconds (Sampling)
## Chain 2:                0.037 seconds (Total)
## Chain 2: 
## 
## SAMPLING FOR MODEL 'anon_model' NOW (CHAIN 3).
## Chain 3: 
## Chain 3: Gradient evaluation took 4e-06 seconds
## Chain 3: 1000 transitions using 10 leapfrog steps per transition would take 0.04 seconds.
## Chain 3: Adjust your expectations accordingly!
## Chain 3: 
## Chain 3: 
## Chain 3: Iteration:    1 / 2000 [  0%]  (Warmup)
## Chain 3: Iteration:  200 / 2000 [ 10%]  (Warmup)
## Chain 3: Iteration:  400 / 2000 [ 20%]  (Warmup)
## Chain 3: Iteration:  600 / 2000 [ 30%]  (Warmup)
## Chain 3: Iteration:  800 / 2000 [ 40%]  (Warmup)
## Chain 3: Iteration: 1000 / 2000 [ 50%]  (Warmup)
## Chain 3: Iteration: 1001 / 2000 [ 50%]  (Sampling)
## Chain 3: Iteration: 1200 / 2000 [ 60%]  (Sampling)
## Chain 3: Iteration: 1400 / 2000 [ 70%]  (Sampling)
## Chain 3: Iteration: 1600 / 2000 [ 80%]  (Sampling)
## Chain 3: Iteration: 1800 / 2000 [ 90%]  (Sampling)
## Chain 3: Iteration: 2000 / 2000 [100%]  (Sampling)
## Chain 3: 
## Chain 3:  Elapsed Time: 0.019 seconds (Warm-up)
## Chain 3:                0.015 seconds (Sampling)
## Chain 3:                0.034 seconds (Total)
## Chain 3: 
## 
## SAMPLING FOR MODEL 'anon_model' NOW (CHAIN 4).
## Chain 4: 
## Chain 4: Gradient evaluation took 2e-06 seconds
## Chain 4: 1000 transitions using 10 leapfrog steps per transition would take 0.02 seconds.
## Chain 4: Adjust your expectations accordingly!
## Chain 4: 
## Chain 4: 
## Chain 4: Iteration:    1 / 2000 [  0%]  (Warmup)
## Chain 4: Iteration:  200 / 2000 [ 10%]  (Warmup)
## Chain 4: Iteration:  400 / 2000 [ 20%]  (Warmup)
## Chain 4: Iteration:  600 / 2000 [ 30%]  (Warmup)
## Chain 4: Iteration:  800 / 2000 [ 40%]  (Warmup)
## Chain 4: Iteration: 1000 / 2000 [ 50%]  (Warmup)
## Chain 4: Iteration: 1001 / 2000 [ 50%]  (Sampling)
## Chain 4: Iteration: 1200 / 2000 [ 60%]  (Sampling)
## Chain 4: Iteration: 1400 / 2000 [ 70%]  (Sampling)
## Chain 4: Iteration: 1600 / 2000 [ 80%]  (Sampling)
## Chain 4: Iteration: 1800 / 2000 [ 90%]  (Sampling)
## Chain 4: Iteration: 2000 / 2000 [100%]  (Sampling)
## Chain 4: 
## Chain 4:  Elapsed Time: 0.02 seconds (Warm-up)
## Chain 4:                0.015 seconds (Sampling)
## Chain 4:                0.035 seconds (Total)
## Chain 4:

2. Check predictions against priors

pp_check(model_priors_only, ndraws = 100)

VARIANCE-COVARIANCE MATRIX

as_draws_df(model) %>%
  select(b_Intercept:sigma) %>%
  cov() %>%
  round(digits = 3)
## Warning: Dropping 'draws_df' class as required metadata was removed.
##                     b_Intercept b_breedckcs b_breedpug b_breedret b_breedother
## b_Intercept               0.090      -0.036     -0.035     -0.037       -0.036
## b_breedckcs              -0.036       0.144      0.035      0.043        0.034
## b_breedpug               -0.035       0.035      0.126      0.035        0.034
## b_breedret               -0.037       0.043      0.035      0.080        0.034
## b_breedother             -0.036       0.034      0.034      0.034        0.066
## sd_visit__Intercept      -0.001      -0.005     -0.002     -0.003        0.000
## sigma                     0.005       0.003      0.000      0.003       -0.001
##                     sd_visit__Intercept  sigma
## b_Intercept                      -0.001  0.005
## b_breedckcs                      -0.005  0.003
## b_breedpug                       -0.002  0.000
## b_breedret                       -0.003  0.003
## b_breedother                      0.000 -0.001
## sd_visit__Intercept               0.132 -0.001
## sigma                            -0.001  0.010

MANUAL POSTERIOR PREDICTIVE DISTRIBUTION CHECKS

NB Uses posterior_predict

1. Posterior predictive distribution plots for breed

# use posterior predict to simulate predictions
ppd <- posterior_predict(model) 

par(mfrow = c(2,2))
stripchart(slgCort ~ breed, vertical = TRUE, method = "jitter",
           col = "steelblue3", data = df, pch = 20, main = "Observed")
stripchart(ppd[sample(seq(1, dim(ppd)[1]), 1),] ~ breed, vertical = TRUE, method = "jitter",
           col = "firebrick3", data = df, pch = 20, main = "PPD")
stripchart(ppd[sample(seq(1, dim(ppd)[1]), 1),] ~ breed, vertical = TRUE, method = "jitter",
           col = "firebrick3", data = df, pch = 20, main = "PPD")
stripchart(ppd[sample(seq(1, dim(ppd)[1]), 1),] ~ breed, vertical = TRUE, method = "jitter",
           col = "firebrick3", data = df, pch = 20, main = "PPD")

ANALYSING THE POSTERIOR DISTRIBUTION

1a. Plot conditional effects from model

plot(conditional_effects(model), ask = FALSE)

1b. advanced plot of conditional effect of breed

ce <- conditional_effects(model, effects = "breed")
ce_df <- ce[[1]][c(1, 6:9)]

ggplot(ce_df, aes(x=breed, y=estimate__, group=1)) +
    geom_errorbar(width=.1, aes(ymin=lower__, ymax=upper__), colour=c("#F8766D", "#A3A500","#00BF7D",
                                                                      "#00B0F6", "#E776F3"), linewidth = 1) +
    geom_point(shape=21, size=6, fill=c("#F8766D", "#A3A500","#00BF7D",
                                        "#00B0F6", "#E776F3")) +
   theme_bw() +
    labs(title = "Conditional effect of breed on hair cortisol") +
         labs(y = paste0("Log Hair Cortisol (standardised)")) +
         labs(x = paste0("breed")) +
         theme(axis.title.y = element_text(size=12, face="bold"), 
               axis.title.x = element_text(size=12, face="bold"),
               title = element_text(size=12, face="bold"),
               plot.title = element_text(hjust = 0.5),
               axis.text.x = element_text(color = "grey25", size = 12),
               axis.text.y = element_text(color = "grey50", size = 10))

2. mcmc_plot of model

a.just parameters of beta variables

mcmc_plot(model,
          variable = c(
         "b_breedckcs",
         "b_breedpug",
         "b_breedret",
         "b_breedother"))

b. all parameters except alpha and sd_visit_intercept

mcmc_plot(model, variable = c(
         "b_Intercept",
         "sigma",
         "b_breedckcs",
         "b_breedpug",
         "b_breedret",
         "b_breedother"))

3. Plot all posterior distributions

posterior <- as.matrix(model)
mcmc_areas(posterior,
pars = c("b_Intercept", "sigma",
         "b_breedckcs",
         "b_breedpug",
         "b_breedret",
         "b_breedother"),
# arbitrary threshold for shading probability mass
prob = 0.75)

4. plot posterior distribution for breeds

posterior <- as.matrix(model)
mcmc_areas(posterior,
    pars = c("b_breedckcs",
         "b_breedpug",
         "b_breedret",
         "b_breedother"),
    prob = 0.75) # arbitrary threshold for shading probability mass

5. Describe the posterior visually

# Focus on describing posterior
hdi_range <- hdi(model, ci = c(0.65, 0.70, 0.80, 0.89, 0.95))
plot(hdi_range, show_intercept = T)

HYPOTHESIS TESTS

1. mix vs. ckcs (from draws) is >0

draws <- as.matrix(model)
mean(draws[,2] >0)
## [1] 0.663375
mean(draws[,2] <0)
## [1] 0.336625

Check 97% credible interval of with HPDI for mix vs. ckcs from draws

HPDI(draws[,2], prob=0.97)
##      |0.97      0.97| 
## -0.7183886  0.9231527

2. mix vs. pug (from draws) is >0

draws <- as.matrix(model)
mean(draws[,3] >0)
## [1] 0.7124167
mean(draws[,3] <0)
## [1] 0.2875833

Check 97% credible interval of with HPDI for mix vs. pug from draws

HPDI(draws[,3], prob=0.97)
##      |0.97      0.97| 
## -0.5865707  0.9615309

2. mix vs. ret (from draws) is >0

draws <- as.matrix(model)
mean(draws[, 4] >0)
## [1] 0.6295417
mean(draws[, 4] <0)
## [1] 0.3704583

Check 97% credible interval of with HPDI for mix vs. ret from draws

HPDI(draws[, 4], prob=0.97)
##      |0.97      0.97| 
## -0.5195171  0.7023931

3. mix vs. other (from draws) is >0

draws <- as.matrix(model)
mean(draws[, 5] >0)
## [1] 0.6539583
mean(draws[, 5] <0)
## [1] 0.3460417

Check 97% credible interval of with HPDI for mix vs. other from draws

HPDI(draws[, 5], prob=0.97)
##      |0.97      0.97| 
## -0.4649280  0.6533254

3. Visualising the posterior of a model using numerical and graphical methods

a. basic (one dog only)

# create new dataframe which contains results of the first dog
new_data <- rbind(df[1,], df[1,], df[1,], df[1,], df[1,])
# Now change one category to be different
new_data$breed <- c("mix", "ckcs", "pug", "ret", "other")
# Visualise df to make sure it has worked
new_data
##   number   group visit season breed_group coat_colour  sex age comorbidity
## 1     c1 stopped    v0 winter         ret        dark Male  43         yes
## 2     c1 stopped    v0 winter         ret        dark Male  43         yes
## 3     c1 stopped    v0 winter         ret        dark Male  43         yes
## 4     c1 stopped    v0 winter         ret        dark Male  43         yes
## 5     c1 stopped    v0 winter         ret        dark Male  43         yes
##   fat_percent cortisol   lgCort breed   slgCort
## 1    52.21393  4.92422 1.594166   mix 0.3415375
## 2    52.21393  4.92422 1.594166  ckcs 0.3415375
## 3    52.21393  4.92422 1.594166   pug 0.3415375
## 4    52.21393  4.92422 1.594166   ret 0.3415375
## 5    52.21393  4.92422 1.594166 other 0.3415375
# Now get mean predictions from the draws of the model
pred_means <- posterior_predict(model, newdata = new_data)


# Compare difference in means for each breedversus mix
differenceCKCS <- pred_means[,1] - pred_means[,2]
differencePug <- pred_means[,1] - pred_means[,3]
differenceRet <- pred_means[,1] - pred_means[,4]
differenceOther <- pred_means[,1] - pred_means[,5]

par(mfrow = c(2,2))

# Examine mean of difference
mean(differenceCKCS)
## [1] -0.1259493
# View histogram of this
hist(differenceCKCS)
# Create HPDI
HPDI(differenceCKCS, 0.93)
##     |0.93     0.93| 
## -2.994134  2.651799
# Examine mean of difference
mean(differencePug)
## [1] -0.1750287
# View histogram of this
hist(differencePug)
# Create HPDI
HPDI(differencePug, 0.93)
##     |0.93     0.93| 
## -2.889368  2.680507
# Examine mean of difference
mean(differenceRet)
## [1] -0.07779497
# View histogram of this
hist(differenceRet)
# Create HPDI
HPDI(differenceRet, 0.93)
##     |0.93     0.93| 
## -2.855377  2.673152
# Examine mean of difference
mean(differenceOther)
## [1] -0.08386029
# View histogram of this
hist(differenceOther)

# Create HPDI
HPDI(differenceOther, 0.93)
##     |0.93     0.93| 
## -2.805863  2.691576

b. Advanced… using all dogs in the model

i. ckcs vs mix

# create new dataframe which contains results of all dogs
new_data1 <- df
# Now change one category to be different
new_data1$breed_group <- c("mix")

# create new dataframe which contains result sof all dogs
new_data2 <- df
# Now change one category to be different
new_data2$breed_group <- c("ckcs")

# Now get predictions from the draws of the models
pred_nd1 <- posterior_predict(model, newdata = new_data1)
pred_nd2 <- posterior_predict(model, newdata = new_data2)
pred_diff <- pred_nd1 - pred_nd2
pred_diff <- data.frame(pred_diff)

# Create mean of differences for each column (dog) of the dataframe
pred_diff_ckcs <- apply(pred_diff, 2, mean)
# View histogram of mean differences
hist(pred_diff_ckcs)

# Examine mean of difference
mean(pred_diff_ckcs)
## [1] -0.001871383
# View histogram of this

HPDI(pred_diff_ckcs, 0.97)
##       |0.97       0.97| 
## -0.01944913  0.01969255

ii. pug vs mix

# create new dataframe which contains results of all dogds
new_data2 <- df
# Now change one category to be different
new_data2$breed_group <- c("pug")

# Now get predictions from the draws of the models
pred_nd1 <- posterior_predict(model, newdata = new_data1)
pred_nd2 <- posterior_predict(model, newdata = new_data2)
pred_diff <- pred_nd1 - pred_nd2
pred_diff <- data.frame(pred_diff)

# Create mean of differences for each column (dog) of the dataframe
pred_diff_pug <- apply(pred_diff, 2, mean)
# View histogram of mean differences
hist(pred_diff_pug)

# Examine mean of difference
mean(pred_diff_pug)
## [1] 0.0005020827
# View histogram of this
HPDI(pred_diff_pug, 0.97)
##       |0.97       0.97| 
## -0.01413850  0.02107792

iii. other vs mix

# create new dataframe which contains results of all dogs
new_data2 <- df
# Now change one category to be different
new_data2$breed_group <- c("other")

# Now get predictions from the draws of the models
pred_nd1 <- posterior_predict(model, newdata = new_data1)
pred_nd2 <- posterior_predict(model, newdata = new_data2)
pred_diff <- pred_nd1 - pred_nd2
pred_diff <- data.frame(pred_diff)

# Create mean of differences for each column (dog) of the dataframe
pred_diff_other <- apply(pred_diff, 2, mean)
# View histogram of mean differences
hist(pred_diff_other)

# Examine mean of difference
mean(pred_diff_other)
## [1] 0.001265779
# Create HPDI
HPDI(pred_diff_other, 0.97)
##       |0.97       0.97| 
## -0.02007916  0.02421905

iv. ret vs mix

# create new dataframe which contains results of all dogs
new_data2 <- df
# Now change one category to be different
new_data2$breed_group <- c("ret")

# Now get predictions from the draws of the models
pred_nd1 <- posterior_predict(model, newdata = new_data1)
pred_nd2 <- posterior_predict(model, newdata = new_data2)
pred_diff <- pred_nd1 - pred_nd2
pred_diff <- data.frame(pred_diff)

# Create mean of differences for each column (dog) of the dataframe
pred_diff_ret <- apply(pred_diff, 2, mean)
# View histogram of mean differences
hist(pred_diff_ret)

# Examine mean of difference
mean(pred_diff_ret)
## [1] -0.0002746355
# Create HPDI
HPDI(pred_diff_ret, 0.97)
##       |0.97       0.97| 
## -0.02206251  0.02043784

Check if better fit if you allow SD to vary arcoss breed

1. Set priors2

# Set individual priors
prior_int <- set_prior("normal(0, 1)", class = "Intercept")
prior_b <- set_prior("normal(0, 1)", class = "b")
prior_sd <- set_prior("normal(0, 2)", class = "sd")
prior_alpha <- set_prior("normal(4, 2)", class = "alpha")

# Combine priors into list
priors2 <- c(prior_int, prior_b, prior_sd, prior_alpha)

2. Run model 2

Increased adapt_delta >0.8 (0.9 here), as had divergent transitions

set.seed(666)
model2 <- brm(bf(slgCort ~ breed + (1 | visit),
                 sigma ~ breed),
                   family = skew_normal(),
                   prior = priors2,
                   data = df,
                   control=list(adapt_delta=0.9999, stepsize = 0.001, max_treedepth =15),
                   iter = 8000, warmup = 2000,
                   cores = 4,
                   save_pars = save_pars(all =TRUE),
                   sample_prior = TRUE)
## Compiling Stan program...
## Trying to compile a simple C file
## Running /Library/Frameworks/R.framework/Resources/bin/R CMD SHLIB foo.c
## using C compiler: ‘Apple clang version 17.0.0 (clang-1700.0.13.5)’
## using SDK: ‘MacOSX15.5.sdk’
## clang -arch arm64 -I"/Library/Frameworks/R.framework/Resources/include" -DNDEBUG   -I"/Library/Frameworks/R.framework/Versions/4.4-arm64/Resources/library/Rcpp/include/"  -I"/Library/Frameworks/R.framework/Versions/4.4-arm64/Resources/library/RcppEigen/include/"  -I"/Library/Frameworks/R.framework/Versions/4.4-arm64/Resources/library/RcppEigen/include/unsupported"  -I"/Library/Frameworks/R.framework/Versions/4.4-arm64/Resources/library/BH/include" -I"/Library/Frameworks/R.framework/Versions/4.4-arm64/Resources/library/StanHeaders/include/src/"  -I"/Library/Frameworks/R.framework/Versions/4.4-arm64/Resources/library/StanHeaders/include/"  -I"/Library/Frameworks/R.framework/Versions/4.4-arm64/Resources/library/RcppParallel/include/"  -I"/Library/Frameworks/R.framework/Versions/4.4-arm64/Resources/library/rstan/include" -DEIGEN_NO_DEBUG  -DBOOST_DISABLE_ASSERTS  -DBOOST_PENDING_INTEGER_LOG2_HPP  -DSTAN_THREADS  -DUSE_STANC3 -DSTRICT_R_HEADERS  -DBOOST_PHOENIX_NO_VARIADIC_EXPRESSION  -D_HAS_AUTO_PTR_ETC=0  -include '/Library/Frameworks/R.framework/Versions/4.4-arm64/Resources/library/StanHeaders/include/stan/math/prim/fun/Eigen.hpp'  -D_REENTRANT -DRCPP_PARALLEL_USE_TBB=1   -I/opt/R/arm64/include    -fPIC  -falign-functions=64 -Wall -g -O2  -c foo.c -o foo.o
## In file included from <built-in>:1:
## In file included from /Library/Frameworks/R.framework/Versions/4.4-arm64/Resources/library/StanHeaders/include/stan/math/prim/fun/Eigen.hpp:22:
## In file included from /Library/Frameworks/R.framework/Versions/4.4-arm64/Resources/library/RcppEigen/include/Eigen/Dense:1:
## In file included from /Library/Frameworks/R.framework/Versions/4.4-arm64/Resources/library/RcppEigen/include/Eigen/Core:19:
## /Library/Frameworks/R.framework/Versions/4.4-arm64/Resources/library/RcppEigen/include/Eigen/src/Core/util/Macros.h:679:10: fatal error: 'cmath' file not found
##   679 | #include <cmath>
##       |          ^~~~~~~
## 1 error generated.
## make: *** [foo.o] Error 1
## Start sampling
## Found more than one class "stanfit" in cache; using the first, from namespace 'rethinking'
## Also defined by 'rstan'
## Found more than one class "stanfit" in cache; using the first, from namespace 'rethinking'
## Also defined by 'rstan'
## Found more than one class "stanfit" in cache; using the first, from namespace 'rethinking'
## Also defined by 'rstan'
## Found more than one class "stanfit" in cache; using the first, from namespace 'rethinking'
## Also defined by 'rstan'
## Warning: There were 1 divergent transitions after warmup. See
## https://mc-stan.org/misc/warnings.html#divergent-transitions-after-warmup
## to find out why this is a problem and how to eliminate them.
## Warning: Examine the pairs() plot to diagnose sampling problems

3. get summary of model

summary(model2)
## Warning: There were 1 divergent transitions after warmup. Increasing
## adapt_delta above 0.9999 may help. See
## http://mc-stan.org/misc/warnings.html#divergent-transitions-after-warmup
##  Family: skew_normal 
##   Links: mu = identity; sigma = log; alpha = identity 
## Formula: slgCort ~ breed + (1 | visit) 
##          sigma ~ breed
##    Data: df (Number of observations: 73) 
##   Draws: 4 chains, each with iter = 8000; warmup = 2000; thin = 1;
##          total post-warmup draws = 24000
## 
## Multilevel Hyperparameters:
## ~visit (Number of levels: 2) 
##               Estimate Est.Error l-95% CI u-95% CI Rhat Bulk_ESS Tail_ESS
## sd(Intercept)     0.61      0.68     0.01     2.54 1.00     6523     9855
## 
## Regression Coefficients:
##                  Estimate Est.Error l-95% CI u-95% CI Rhat Bulk_ESS Tail_ESS
## Intercept           -0.07      0.48    -1.10     0.95 1.00     7882     8794
## sigma_Intercept      0.05      0.19    -0.30     0.45 1.00    10816    13304
## breedckcs            0.05      0.41    -0.73     0.92 1.00    13361    13495
## breedpug             0.26      0.48    -0.64     1.27 1.00    13708    13847
## breedret            -0.02      0.32    -0.66     0.61 1.00    13048    14860
## breedother           0.15      0.32    -0.48     0.76 1.00    13071    14635
## sigma_breedckcs     -0.11      0.36    -0.77     0.64 1.00    12868    14771
## sigma_breedpug       0.16      0.35    -0.48     0.90 1.00    13047    15225
## sigma_breedret      -0.11      0.27    -0.63     0.41 1.00    12526    15927
## sigma_breedother     0.04      0.24    -0.44     0.50 1.00    12516    14554
## 
## Further Distributional Parameters:
##       Estimate Est.Error l-95% CI u-95% CI Rhat Bulk_ESS Tail_ESS
## alpha     4.17      1.39     1.86     7.24 1.00    18009    14160
## 
## Draws were sampled using sampling(NUTS). For each parameter, Bulk_ESS
## and Tail_ESS are effective sample size measures, and Rhat is the potential
## scale reduction factor on split chains (at convergence, Rhat = 1).

4. Try the PSIS LOO-CV procedure to check model performance

loo_model2 <- loo(model2, moment_match = TRUE)
loo_model2
## 
## Computed from 24000 by 73 log-likelihood matrix.
## 
##          Estimate   SE
## elpd_loo   -109.6  7.2
## p_loo        11.2  2.6
## looic       219.1 14.3
## ------
## MCSE of elpd_loo is 0.0.
## MCSE and ESS estimates assume MCMC draws (r_eff in [0.5, 1.1]).
## 
## All Pareto k estimates are good (k < 0.7).
## See help('pareto-k-diagnostic') for details.

5. Compare looic for models 1 and 2

model <- add_criterion(model, "loo")
## Warning: Found 1 observations with a pareto_k > 0.7 in model 'model'. We
## recommend to set 'moment_match = TRUE' in order to perform moment matching for
## problematic observations.
model2 <- add_criterion(model2, "loo")
## Warning: Found 3 observations with a pareto_k > 0.7 in model 'model2'. We
## recommend to set 'moment_match = TRUE' in order to perform moment matching for
## problematic observations.
loo_compare(model, model2)
##        elpd_diff se_diff
## model   0.0       0.0   
## model2 -4.4       2.0

Model 1 is a better fit so keep this